def transform(dataframe: pd.DataFrame,
scaler: TransformerMixin) -> pd.DataFrame:
fields_to_normalize = dataframe.filter(
['preco', 'prazo', 'frete', 'latitude', 'longitude']).to_numpy()
feature_scaled = scaler.fit_transform(fields_to_normalize)
dataframe['features'] = list(feature_scaled)
return dataframe
def _fit_transform_with_state_restore_check(transformer: TransformerMixin, X,
**kwargs):
transformed = transformer.fit_transform(X, **kwargs)
LOGGER.debug('transformed: %s', transformed)
LOGGER.debug('transformed.shape: %s', transformed.shape)
restored_transformer = _get_state_and_restore(transformer)
restored_transformed = restored_transformer.transform(X)
LOGGER.debug('restored_transformed: %s', restored_transformed)
assert restored_transformed.tolist() == transformed.tolist()
return transformed
def produce_fit(kmeans_alg, features: np.ndarray, transformer: TransformerMixin):
start_time = perf_counter()
# dim reduction
features_transformed = transformer.fit_transform(features)
# K-means
return kmeans_alg.fit_predict(features_transformed), perf_counter() - start_time
def scale_data(dataset: array, scaler: TransformerMixin) -> array:
assert scaler is not None and dataset is not None
return scaler.fit_transform(dataset)
